IB/qib: Optimize locking for get_txreq()

The current code locks the QP s_lock, followed by the pending_lock, I
guess to to protect against the allocate failing.

This patch only locks the pending_lock, assuming that the empty case
is an exeception, in which case the pending_lock is dropped, and the
original code is executed.  This will save a lock of s_lock in the
normal case.

The observation is that the sdma descriptors will deplete at twice the
rate of txreq's, so this should be rare.

Signed-off-by: Mike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

drivers/infiniband/hw/qib/qib_verbs.c

@@ -913,8 +913,8 @@ static void copy_io(u32 __iomem *piobuf, struct qib_sge_state *ss,
 		__raw_writel(last, piobuf);
 }
 
-static struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
-					 struct qib_qp *qp, int *retp)
+static noinline struct qib_verbs_txreq *__get_txreq(struct qib_ibdev *dev,
+					   struct qib_qp *qp)
 {
 	struct qib_verbs_txreq *tx;
 	unsigned long flags;
@@ -926,8 +926,9 @@ static struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
 		struct list_head *l = dev->txreq_free.next;
 
 		list_del(l);
+		spin_unlock(&dev->pending_lock);
+		spin_unlock_irqrestore(&qp->s_lock, flags);
 		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
-		*retp = 0;
 	} else {
 		if (ib_qib_state_ops[qp->state] & QIB_PROCESS_RECV_OK &&
 		    list_empty(&qp->iowait)) {
@@ -935,14 +936,33 @@ static struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
 			qp->s_flags |= QIB_S_WAIT_TX;
 			list_add_tail(&qp->iowait, &dev->txwait);
 		}
-		tx = NULL;
 		qp->s_flags &= ~QIB_S_BUSY;
-		*retp = -EBUSY;
+		spin_unlock(&dev->pending_lock);
+		spin_unlock_irqrestore(&qp->s_lock, flags);
+		tx = ERR_PTR(-EBUSY);
 	}
+	return tx;
+}
 
-	spin_unlock(&dev->pending_lock);
-	spin_unlock_irqrestore(&qp->s_lock, flags);
+static inline struct qib_verbs_txreq *get_txreq(struct qib_ibdev *dev,
+					 struct qib_qp *qp)
+{
+	struct qib_verbs_txreq *tx;
+	unsigned long flags;
 
+	spin_lock_irqsave(&dev->pending_lock, flags);
+	/* assume the list non empty */
+	if (likely(!list_empty(&dev->txreq_free))) {
+		struct list_head *l = dev->txreq_free.next;
+
+		list_del(l);
+		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		tx = list_entry(l, struct qib_verbs_txreq, txreq.list);
+	} else {
+		/* call slow path to get the extra lock */
+		spin_unlock_irqrestore(&dev->pending_lock, flags);
+		tx =  __get_txreq(dev, qp);
+	}
 	return tx;
 }
 
@@ -1122,9 +1142,9 @@ static int qib_verbs_send_dma(struct qib_qp *qp, struct qib_ib_header *hdr,
 		goto bail;
 	}
 
-	tx = get_txreq(dev, qp, &ret);
-	if (!tx)
-		goto bail;
+	tx = get_txreq(dev, qp);
+	if (IS_ERR(tx))
+		goto bail_tx;
 
 	control = dd->f_setpbc_control(ppd, plen, qp->s_srate,
 				       be16_to_cpu(hdr->lrh[0]) >> 12);
@@ -1195,6 +1215,9 @@ unaligned:
 	ibp->n_unaligned++;
 bail:
 	return ret;
+bail_tx:
+	ret = PTR_ERR(tx);
+	goto bail;
 }
 
 /*